Marketing posters and feed graphics
Put the exact headline in the prompt to have it rendered crisply. Afterwards, separate the poster into layers to move the wording or replace it with another language.
Loading Seedream 5.5...
One model covers three jobs here: producing new pictures from a written prompt, revising pictures you upload, and separating a finished picture into layers. Edits can be stacked without the subject changing, and backgrounds can be left transparent. An edit accepts as many as ten references; layer separation takes a single upload and returns a base picture with PNG layers on clear backgrounds. Output is 1K, 1.5K or 2K, and the price in credits is displayed before each run.
The model is a good fit when the job calls for reusable assets, a subject that stays the same from picture to picture, or lettering that can actually be read.
Put the exact headline in the prompt to have it rendered crisply. Afterwards, separate the poster into layers to move the wording or replace it with another language.
Produce a cut-out of the product on a clear background, then drop it into holiday or campaign scenes through editing; the packaging text and outline are preserved.
Design a mascot or hero once, feed it back as a reference, and place it in different poses, clothes and locations for comic strips, game art or brand campaigns.
Request single elements on a clear background for use in slide decks, mobile apps, merchandise printing and chat sticker packs.
Take out an unwanted object, shift the lighting or the weather, or swap the backdrop in a sequence of small steps; everything you leave unmentioned stays as it was.
Convert a flattened banner or photograph into separate layers and continue the work in Figma, Photoshop or another layered editor, with no need to redraw it.
The mode switch sits above the form. Settings are remembered per mode, which makes it easy to generate a picture, revise it and then separate it into layers in one sitting.
Prompts can be written in whichever language you prefer. Set the shape (1:1 through 21:9), the size (1K, 1.5K or 2K) and the background. A Transparent background gives an isolated subject as a clear PNG; a Solid background gives a full scene.
Add between one and ten references and state what should be different, for example a new backdrop, another product color, a character moved into a different setting or a new visual style. Leave the shape on Auto to follow the reference, or choose another.
Provide a single picture. Optionally list what should be pulled apart (a person, a logo, a headline) and enable text separation so detected wording gets layers of its own. The base picture comes first in the results, then the PNG layers with clear backgrounds.
All three Seedream models can be selected on this site. Seedream 5.5 is the choice for stacked edits, transparent output and layers. Seedream 5 Pro and Seedream 5 Lite remain worthwhile if your prompts are already tuned to them.
| Capability | Seedream 5.5 | Seedream 5 Pro | Seedream 5 Lite |
|---|---|---|---|
| Available modes | Generate, edit and layers | Generate and edit | Generate and edit |
| References per edit | 10 maximum | 10 maximum | 10 maximum |
| PNG with transparency | Yes | No | No |
| Layer separation | Yes | No | No |
| Output sizes | 1K, 1.5K, 2K | 1K, 2K | 2K, 4K |
This release concentrates on three things: edits that can be repeated safely, output that can be reused as assets, and lettering that stays legible. Each item below is available in the form on this page.
With the background set to Transparent, the PNG carries a true alpha channel. Cut-outs of products, logos, icons and stickers can go straight onto a layout, and no extra background remover is required.
Adjust a single detail, then the next. Areas outside your instruction are left alone from round to round, which keeps faces, products and scenery from shifting during a long revision. A small item such as glasses or a ring can still be altered accurately after up to 100 rounds.
A single upload comes back as a base picture and individual PNG layers, with clear backgrounds, for the people, objects and wording it contains. In a design tool, every layer can be repositioned, recolored or swapped out.
Titles, product labels, menus and fine print are rendered more cleanly, with fewer malformed characters. Posters, packaging concepts and social posts that depend on readable wording benefit most.
Earlier outputs can serve as references, so one character, mascot or type style carries through a set of pictures, whether that is one portrait or an entire campaign.
Textures such as skin, cloth, glass, metal and food are lit more believably. Prompts written like a photo brief look closer to studio work and require less retouching afterwards.
A single run takes only a few minutes from start to download. The steps below apply to every run in the form at the top of the page.
Text to Image creates a new picture, Image Editing revises pictures you already have, and Layers breaks one picture into its parts.
State the subject, the setting, the light and the style. When editing, attach references and spell out both the change and the parts that must remain.
Select shape, size, background and file format. The credit price for those settings is printed on the Generate button.
Most results are ready in less than a minute. Save them as PNG or JPEG, or press Edit Again to reuse a result as the reference for another round.
Short answers on pricing, modes, output and troubleshooting.
Use our AI image prompt gallery for creative inspiration, then generate them with Seedream 5.5.
Browse AI image prompts →Loading Seedream 5.5...